Reduced lung function is independently associated with increased risk of type 2 diabetes in Korean men

Study subjects were selected from a total of 10,965 men who visited Kangbuk Samsung Hospital Health Promotion Center for health examinations in both 2003 and 2008. The purpose of the medical health checkup program is to promote the health of the employees through regular health checkups and to enhance early detection of existing diseases, if any. Most of the examinees are the employees and family members of various industrial companies from all around the country. The costs of the medical examinations are largely paid for by their employers, and a considerable proportion of the examinees undergo examinations annually or biannually. We took advantage of this opportunity to conduct a follow-up study.

Among the potential study subjects, 1,745 individuals were excluded; 139 individuals had a history of type 2 diabetes, 237 individuals had fasting plasma glucose ≥ 7.0 mmol/L (126 mg/dl) in 2003 (baseline visit), 31 individuals did not have available FEV₁ or FVC level measurements, and 1,338 individuals had no available data regarding education, alcohol, smoking or exercise. Finally, 9,220 men (mean age: 41.4 years, range: 24 to 82 years) who did not have type 2 diabetes in 2003 (baseline visit) were enrolled in the study and were followed up for five years. The study subjects were divided into quartiles according to the baseline percentage of predicted values (% predicted) for FEV₁ or FVC. Based on FEV₁ (% predicted), the resulting four categories were as follows: ≤94.6% in quartile 1, 94.6-105.5% in quartile 2, 105.5-119.9% in quartile 3, and >119.9% in quartile 4. The subjects were similarly divided into quartiles based on FVC (% predicted): ≤89.4% in quartile 1, 89.4-98.6% in quartile 2, 98.6-109.1% in quartile 3, and ≥109.1% in quartile 4. We then compared the incidence of type 2 diabetes according to different FVC (% predicted) and FEV₁ (% predicted) groups after a follow-up survey.

Medical and medication history, smoking status (current-, ex-, or non-smoker), alcohol consumption (g/day), regular exercise (≥ once per week), and education status were assessed using the same standard questionnaire in 2003 and 2008. Blood pressure (BP) was measured with a standard sphygmomanometer following at least five minutes of seated rest. Height and weight were estimated using automated instruments, with individuals wearing light clothing and no shoes. Body mass index (BMI) was calculated as body weight (kilograms) divided by height squared (meters²).

The study protocol conformed to ethical guidelines of the 1975 Declaration of Helsinki, and accordingly the Kangbuk Samsung Hospital Human Research Committee approved it. The Kangbuk Samsung Hospital Institutional Review Board also approved this study and each participant gave their written informed consent.

Morning blood samples were drawn from the antecubital vein after participants had fasted for at least 12 hours. Plasma glucose levels were determined using the hexokinase method (Advia 1650 AutoAnalyzer, Bayer Diagnostics, Leverkusen, Germany). Insulin concentrations were measured using immunoradiometric assays (RIABEAD II, Abbott, Tokyo, Japan), with intra- and inter-assay coefficients of variance of 1.2-1.9% and 1.4-3.3%, respectively. Insulin resistance was estimated using the homeostasis model assessment of insulin resistance index (HOMA-IR), which is defined as fasting insulin (μIU/ml) x fasting glucose (mmol/L)/22.5 [16]. Fasting total cholesterol (TC) and triglyceride (TG) levels were measured enzymatically using an automatic analyzer (Advia 1650 AutoAnalyzer, Bayer Diagnostics, Leverkusen, Germany). High density lipoprotein cholesterol (HDL-C) concentrations were measured using a selective inhibition technique. Levels of low density lipoprotein cholesterol (LDL-C) were determined through a homogeneous enzymatic calorimetric test. Type 2 diabetes was defined as fasting plasma glucose ≥ 7.0 mmol/L (126 mg/dL) based on 1997 American Diabetes Association criteria, current usage of diabetes medication, or positive response to the question, “Has a medical person ever told you that you had diabetes?” [17].

Spirometry was performed as recommended by the American Thoracic Society [18] using V_max22 (SensorMedics, Yorba Linda, CA, USA). Absolute values of FEV₁ and FVC were obtained, and the percentage predicted values (% predicted) for FEV₁ and FVC were calculated from the following equations obtained in a representative Korean population sample [19]:

The highest FEV₁ and FVC values from three or more tests with acceptable curves were used for further analyses. The FVC (% predicted) and FEV₁ (% predicted) were calculated by dividing the FVC (L) and FEV₁ (L) by the predicted FVC and FEV₁, respectively.

The reliability of the spirometry is proved in the previously published report [15].

Data are expressed as the mean ± standard deviation (SD) for continuous variables and as percentages for categorical variables. Serum TG concentration was log-transformed for analysis to correct skewed distributions, but the values in the tables are expressed as untransformed data for easy interpretation. Comparisons of baseline cardiovascular risk factors according to the presence/absence of incident type 2 diabetes were made using Student’s t-test for continuous variables or the chi-square test for categorical variables. Comparisons of baseline variables between the lowest and highest quartiles of FEV₁ (% predicted) or FVC (% predicted) were made by Student’s t-test or the chi-square test. Comparisons of type 2 diabetes development according to FEV₁ (% predicted) or FVC (% predicted) quartiles were obtained from chi-square tests. Multivariable logistic regression analyses were conducted to assess the relationship between FEV₁ (% predicted) or FVC (predicted) quartiles and the risk of incident type 2 diabetes: model 1 was adjusted for age and BMI; model 2 was adjusted as in model 1 plus education, smoking, exercise, alcohol, and insulin; model 3 was adjusted as in model 1 plus education, smoking, exercise, alcohol, and HOMA-IR. The outcome in this study was type 2 diabetes at five years and all subjects had been followed for approximately five years. Therefore, our data for the analysis was complete (not censored) in terms of outcome. For this reason, data was analyzed by logistic regression instead of using a Cox regression model. All statistical analyses were performed using PASW for Windows, version 18.0 (SPSS Inc., Chicago, IL, USA). All statistical tests were two-tailed, and P-values < 0.05 were considered statistically significant.

Baseline characteristics of the non-diabetes group and the diabetes group at follow-up are presented in Table 1. Individuals in the diabetes group were older and more likely to have higher mean BP, heart rate (HR), BMI, TC, TG, glucose, insulin, and HOMA-IR, and lower HDL-C, FEV₁ (L), and FVC (L) values compared with those in the non-diabetes group. In addition, there were more subjects in the diabetes group who were obese, currently smoking, and who had a low-level of education (Table 1).

The baseline characteristics of the lowest and highest quartiles of FEV₁ (% predicted) and FVC (% predicted) were compared (Table 2). In the lowest quartile (quartile 1) of FEV₁ (% predicted), TG, glucose and the percentage of current smokers were significantly higher compared to the highest quartile (quartile 4). However, age, HR, BMI, HDL-C, insulin and alcohol consumption were significantly lower in the lowest quartile compared to the highest quartile. On the other hand, subjects in quartile 1 of FVC (% predicted) were older, more likely to have higher BP, BMI, TG, glucose, insulin and HOMA-IR, and had lower HR and HDL-C compared to quartile 4.

In age- and BMI-adjusted logistic regression analysis (model 1), the lowest quartiles of FVC (% predicted) and FEV₁ (% predicted) had a higher odds ratio (OR) for the incidence of type 2 diabetes compared to the highest quartile reference category (OR [95% confidence interval (CI)], 2.18[1.42-3.32] and 1.95[1.31-2.92], P < 0.0001 and P = 0.001, respectively) (Table 3, Figure 1). The OR results from models 2 and 3 were attenuated, but were still significant in FVC (% predicted) and FEV₁ (% predicted). Moreover, analysis with model 3 including age, BMI, education, smoking, exercise, alcohol, and HOMA-IR revealed that the ORs in FVC (% predicted) and FEV₁ (% predicted) quartile 1 were significantly higher compared with those in quartile 4 (OR [95% CI], 1.90[1.23-2.93] and 1.66[1.10-2.50], P = 0.004 and P = 0.019, respectively).

We also conducted a sub-analysis of 5,417 non-obese subjects with BMI <25 (mean BMI: 22.7 ± 1.6, range: 16.1 to 24.9) to evaluate the association between lung function and type 2 diabetes in the non-obese population. In the non-obese subjects with BMI <25, quartile 1 of FVC (% predicted) and FEV₁ (% predicted) had a significantly increased OR for type 2 diabetes compared to quartile 4 after adjusting for age and BMI (OR[95% CI], 2.15[1.02-4.57] and 2.19[1.09-4.42], P = 0.045 and P = 0.028, respectively) (Table 4).

We note that our study had some limitations. First, there is the possibility of selection bias because most participants were residents of an urban community, all subjects were of Korean descent, from one university hospital, and there were no women in this study. Therefore, the results of this study cannot be generalized to the worldwide population. Second, incident type 2 diabetes was defined based only on fasting glucose level or a self-reported medical history of type 2 diabetes, and no oral glucose tolerance tests were performed. Thus, the incidence of type 2 diabetes may have been underestimated in this study. Third, risk of diabetes may have been influenced by early life style factors which was not investigated in this study. Therefore, future studies including birth cohort analysis are needed to clarify the cause-effect link between reduced lung function and risk of developing diabetes. Finally, our data consisted of only baseline and five-year follow-up time points. The outcome was considered the presence of type 2 diabetes at five years. Thus, logistic regression was used for analysis instead of a Cox regression model. However, our study had the advantage of including a relatively large total number of subjects over a relatively long follow-up period.